Deep Learning-Powered AI System for Automated Detection of Common Ocular Diseases from Medical Images
Main Article Content
Abstract
Glaucoma, cataracts, diabetic retinopathy, and age-related macular degeneration are among the most common eye disorders globally, although numerous others can also lead to vision loss or impairment. Despite its critical importance, early detection is difficult to achieve because of the sensitivity of the symptoms and the need for skilled interpretation of medical pictures. Using fundus images from the ODIR dataset, this work suggests an AI system powered by deep learning for automated identification of common eye illnesses. A multi-class classification utilizing a fine-tuned VGG-16 convolutional neural network is part of the technique, along with data augmentation to rectify class imbalance and preprocessing methods like CLAHE and Gaussian denoising. The model achieved impressive classification performance, with a recall of 90.00%, an F1-score of 92.00%, a precision of 94.00%, and an accuracy of 98.27%. Examining the suggested model in comparison to DenseNet121, CNN, and VGG-16 + CNN reveals that it offers a better equilibrium between recall and precision. In terms of scalable and dependable eye disease screening, these findings validate the efficacy and therapeutic promise of the suggested method. In addition to highlighting the significance of deep learning integration in medical imaging workflows, the paper proposes future improvements that involve attention mechanisms and ensemble learning to increase sensitivity.
Downloads
Article Details
Section

This work is licensed under a Creative Commons Attribution 4.0 International License.